Abstract: Efficient and accurate calibration is the key issue for 3D model reconstruction based on the depth image．Based on the synchronous color image and depth image obtained from the Kinect camera,a calibration method based on the sparse feature point was presented．We took the color image as a reference image,used the geometric relation between the depth image and the color image to calculate the corresponding matrix transformation parameters．Then we extracted corresponding sparse feature points of the two types of data to determine the calibration parameters,and thus obtain a mapping relationship between the two data．Finally,we achieved the calibration of depth information according to the determined parameters．The results show that the relative transformation between the depth image and the color image meets a linear approximation conversion．This method has low computational complexity,high accuracy and easy to realize．So it is suitable for further promotion in related fields.

Key words: Depth image,Image calibration,Sparse feature points,Feature points matching,Machine vision